Speech-dispatcher c api questions

[Luke Yelavich]

On Tue, Jun 16, 2009 at 04:14:20PM EST, Jeremy Whiting wrote:
> Hynek, others,
> 
> I've been using the Speech-dispatcher c api for a bit integrating it into 
> kttsd to replace a lot of functionality in kttsd itself as you know. =)  
> Anyway, I came across a couple things I wanted to ask about on the list.
> 
> 1) I notice after some time kttsd still gets job numbers back from calling 
> say() but speech-dispatcher stops sending audio out to the speakers.  I'm 
> using espeak and alsa but haven't figured out where the hang up is yet (I 
> noticed once today speechd was using 50% of my cpu at the time.  Is this 
> known? I'll try to get using festival tomorrow and see how that works.

There is a bug in the alsa output code that has yet to be found, causing the 
sd_espeak process to segfault, and yes, speech-dispatcher keeps running. SO 
this is known, but nobody has yet found a fix for it.

> 2) I'm using one connection for the whole lifecycle of kttsd (unless it goes 
> down, then I reconnect.  Is this the preferred method? or should I be making 
> a 
> new connection for each speech job?

I would think 1 connection for the lifecycle is preferred. I think even Orca 
does this as well.

> 3) When I set kttsd up to send all irc notifications through 
> speech-dispatcher 
> I notice that jobs are getting truncated.  I'm currently saying all jobs with 
> a text priority and it seems when a new job comes in, the previous job is 
> cancelled and the new one started.

That sounds right, probably similar to what orca does. When orca sends text to 
speak, and the user presses a key to read another line, the previous job is 
cancelled. I dare say kttsd will have to cue the jobs, and play the new 
notification once the other has finished.

> I think that's all for now (quite a bit I know) anyway, thanks for the help 
> and I'll be glad to hack on speech-dispatcher itself if needed.  I cloned the 
> git repository this morning (not sure how pushes go with this setup, I've 
> done 
> merge requests via gitorious before).

Bare in mind that the repo doesn't yet have authorship info for commits, which 
I hope will be rectified soon. As for commit access, you will have to ask for 
an account to be able to commit to the speech-dispatcher git repo.

Luke